Figure 3.1 The EGG - Elastic General-purpose Grip
3.2.1 Experiment 2 Set-up
3.2.1.1 Experimental Platform, Task, and Display
The experimental platform, task, and display in this experiment
are the same as those used in Experiment 1. A preliminary analysis
of this experiment was published in (Zhai, 1993).
3.2.1.2 Controllers Used in the Experiment - the Design of the EGG
In this experiment, the isometric rate control mode was implemented by means of a Spaceballô.
In order to carry out research on isometric and elastic 6 DOF input, the EGG (Elastic General purpose Grip), a 6 DOF elastic device, was designed, as shown in Figure 3.1.
The EGG was implemented with an egg shaped handle suspended in a 17 cm x 17 cm x14 cm frame by eight sets of elastic bands joined to each corner of the frame. The handle, with horizontal diameter 6.5 cm, same as that of a Spaceball (model 2003), and vertical diameter 8 cm, is grasped by fingers in the gesture of a Ïprecision gripÓ (MacKenzie and Iberall, 1994) . An electromagnetic tracker (the Ascension Bird) is mounted in the centre of the handle for acquiring the 6 DOF data. The elastic suspension was arranged such that the grip did not bias the manipulation in any particular direction. In comparison with other proposed elastic 6 DOF designs (e.g. McKinnon et al., 1987) , where the handle is either suspended from the ceiling or supported on a base, the EGG provides a substantially uniform compliance in all directions for both rotational and translational movement, thereby allowing control actions to be decoupled. Another decoupled elastic 6 DOF design is described by Hayward, Nemri, Chen, and Duplat (1993) , in which a mechanically much more sophisticated design is proposed.
The EGG allows movement of 20 mm in translations and 30_ in rotations,
which are much greater than the currently available commercial
6 DOF devices (see Figure 1.1).
Another 6 DOF elastic controller prototype was also developed, as illustrated in Figure 3.2. It had a larger frame and a bar handle to be grasped with the gesture of a "power grip" (MacKenzie and Iberall, 1994) . The EGG was chosen for the present experiment, however, because its handle has dimensions similar to those of the Spaceball (Model 2003). In comparison to the design in Figure 3.2, the EGG also has the advantage of allowing utilisation of the fine muscle groups (fingers) instead of the wrist, elbow and shoulder alone. Neurological studies have shown that the fingers have a rich representation in the somatosensory and motor cortex. In Chapter 4, a separate study is presented to test the advantages of using fingers in 6 DOF manipulation.
The structure of the EGG also made it very simple to adjust the elastic stiffness, since the number of suspending elastic bands could be easily changed. As discussed before, the optimal elasticity for such as device is a result of the trade-off between compatibility with rate control (requiring stiff loading) and proprioception (requiring loose loading). Indeed, a pilot study with the EGG showed that both extremes of elasticity settings produced poorer performance. Optimal performance was found when the elasticity was around 120g/cm in each of the X, Y, Z directions.
3.2.1.3 Models of the Input Controls
Both the isometric rate control and the elastic rate control have the same mathematical model as the isometric rate control model described in Appendix 1, section A1.4.
3.2.1.4 Optimisation of Control Gains
The control gains (sensitivities) for each condition were optimised
through systematic parameter searching, as described in section
2.7. For the isometric rate controller (Spaceball), the same gain
selected for Experiment 1 was used here (see Figure 2.7). U shaped
performance-gain curve was also found for the EGG (Figure 3.3).
The elasticity of the elastic rate controller was optimised in
a similar fashion. In order to equalise the operating conditions
of the two controllers as much as possible, the non-linear transformations
of the translation and rotation degrees of freedom embedded in
the Spaceball, as described in section 2.7 of Chapter 2, were
also applied to output variables from the EGG.
3.2.2 Experimental Design
3.2.2.1 Subjects
A between subjects design was employed in this experiment. Each of the subjects served in only one condition: isometric rate control or elastic rate control. One of the pitfalls of between-subjects designs is that individual differences may bias experimental results. Pitrella and Kruger (1983) have suggested using matching tests to form equal groups for tracking experiments. However, choosing a suitable matching test is a very delicate task, since the test has to be sufficiently similar to the experimental conditions that measured and matched subjects' capabilities will be relevant to the experimental task. On the other hand, the test also has to be such that the amount of skill transferred from the matching test to each of the experimental conditions is equal, so that the matching test does not introduce a bias into the actual experimental results. It is often impossible to design such a test to fit all these requirements.
In this experiment, randomisation and a relatively large number of subjects were used to dilute any possible individual difference effects. 35 paid volunteer subjects were recruited by advertising through posters and electronic network news groups on the University of Toronto campus. People who participated in Experiment 1 were excluded from this experiment. All subjects were screened using a Bausch and Lomb Orthorater. Five of the subjects were rejected for having poor (corrected) near vision acuity. Another four were rejected for having weak stereo-acuity.
Among the 26 subjects accepted, two were left-handed, as determined by the Edinburgh Inventory (Oldfield, 1971). One of them was assigned to the elastic rate control condition and the other was assigned to the isometric rate control condition. The controls were set at the side of the subjects' dominant hand. Three of the 26 accepted subjects were female. Two were put into the elastic rate control condition and one was assigned to the isometric rate condition. The remaining 21 male right handed subjects were randomly assigned to the two conditions. The balance of composition of the two groups of subjects was checked by age, profession, etc. No obvious bias against any condition could be found.
The accepted subjects' ages ranged from 16 to 38, with the majority
in their early to mid-20's. Most of the subjects were engineering
or computer science undergraduate students. All had experience
with computer mice but none of them had used a 6 DOF input device
before the experiment.
3.2.2.2 Experimental Procedure
Each experimental session consisted of a 10 minute vision screening
test, 5 minutes of instruction, 40 minutes of experiment, and
a 5 minute questionnaire survey. The 40 minute experiment was
divided into four phases. Each phase comprised 10 minutes of training,
followed by 12 trials of data collection. Each training phase
consisted of demonstrations and coaching by the experimenter,
combined with practice trials. The data from the 12 trials were
composed of 3 blocks of 4 trials, each block comprising 4 different
randomly shuffled starting locations for the manipulated tetrahedron
(the cursor). The procedure was the same as Experiment 1, except
that Experiment 1 had a within subject design and this had a between
subject design.
3.2.3 Results
3.2.3.1 Performance Results
Figure 3.4 displays the means and standard errors of the two techniques over the four phases of the experiment. The results of a repeated measure variance analysis of the entire data set are summarised in Table A3.2.1. The performance difference between the two input techniques was not statistically significant: F(1, 24) < 1. However, the Phase x Input interaction was weakly significant: F(3, 72) = 2.57, p = 0.061. This means that the relative performance with each of the two input techniques is likely related to learning experience. As shown in Figure 3.4, on average the elastic rate control outperformed the isometric rate control in Phase 1. In later phases, however, the difference between the two was reduced to practically zero.
Learning was therefore the most important factor affecting performance.
Subjects' performances improved significantly with practice for
both input techniques: F(3, 72) = 90.5, p < 0.0001. The initial
target location was also significant in affecting trial completion
time: F(3, 72) = 4.5, p < 0.01. This was because two of the
initial locations were much farther from the target than the other
initial locations.
3.2.3.2 Subjective Ratings
Immediately after each session of the experiment, the subjects
were asked to comment on the ease of use/ difficulty and the degree
of fatigue for the controller that they had just used. The subjective
opinions between the two controllers were not statistically different.
For ease of use: F(1, 24) < 1, p > 0.5, and for fatigue,
F(1, 24) < 1, p > 0.5 (See Figures 3.5 and 3.6).
On the average across the four experimental phases, the results of Experiment 2 did not show substantial differences between the two controllers used in the experiment. It therefore did not provide strong evidence for either the superiority of elastic devices, as suggested by researchers such as Poulton, or the superiority of isometric devices, as suggested by researchers such as Gibbs. Interestingly, the relative advantage of the isometric and the elastic device appears to be related to learning, as reflected by the interaction between experimental phase and control technique, since a modest advantage for the elastic device was shown in the first but not in the later experimental phases.
The 6 DOF docking task used in this experiment was a relatively
easy task and thus performance ceiling effects could have concealed
the differences between the two modes. It was therefore decided
to further test the two types of controllers by conducting a more
demanding task - 6 DOF tracking. Preliminary results of the tracking
experiment were published in (Zhai and Milgram 1993b) and (Zhai
and Milgram 1994b). A more complete description of the experiment
is presented in the following section.